The roughly US$2 trillion ($2.8 trillion) sell-off in the global software sector since September 2025 is more than a painful drawdown for growth investors. It is a live stress test of how portfolios behave when a “quality compounder” sector is re-priced for structural disruption rather than interest rates.

For asset owners, the implication is immediate: this is less a call on one sector than a test of portfolio construction and governance. “Software” exposure often sits embedded across growth managers, quality tilts, and private market sleeves, so the priority is to identify hidden concentrations, stress-test underlying business-model risk, and ensure active managers are being paid for genuine selectivity rather than style.

Two forces are colliding – and they transmit differently through multi-manager portfolios. Cyclically, hyperscalers’ AI infrastructure spend is crowding out enterprise IT budgets, redirecting dollars from application software towards compute and data centre capacity. Structurally, AI agents threaten seat-based SaaS economics by automating tasks and reducing the number of paid human users required to produce the same output.

Notably, this repricing has happened without a broad market panic. That is precisely why it matters: when headline indices look calm, investors can miss the build-up of stock-specific dispersion and hidden concentration risk inside portfolios.

Headline volatility measures such as the VIX rose only modestly as the sell-off gathered pace. But those averages mask widening outcome ranges within software, and increase the value of research depth, risk discipline and genuine selective exposure.

To understand the industry outlook, investors must first grasp the important nuances in SaaS business models. These distinctions mean a number of businesses retain core strengths, such as recurring revenue and high margins, despite the risk of AI disruption.

For example, AI has the greatest impact on ‘seat-based’ revenue for user interface tool software businesses. This is the model where customers pay a recurring fee for a set number of individual users completing workflows that sit over customer data.

AI’s impact is less for deeply embedded and mission-critical systems of record, as they are more difficult to replace. Company employee payroll software is one example that falls into this category.

Bulls vs bears

Both overly bearish and overly bullish views about AI miss such critical differences in its likely adoption and likely evolution of revenue streams. They also fail to reflect the manner in which businesses may respond.

SaaS bulls tend to dismiss AI adoption as a cyclical budget rotation, ignoring the structural shifts within the industry and the progression of AI capability. They may also underestimate the impact on seat-based revenue as AI reduces human headcount and the shift from per-seat pricing to outcome or task-based pricing models.

There is a reasonable argument to suggest that the framework used for valuing software companies needs to be more conservative. This is because company earnings multiples may not re-rate to historical averages, given the predictability of cash flows and margins are under scrutiny.

Meanwhile, the industry bears overestimate the speed and scale of AI agent replacement in enterprise environments. They may also under-appreciate the complexities of compliance, integration and data governance. Other factors they ignore include the lengthy process to replace established SaaS platforms with AI alternatives – instead misinterpreting early AI demos as immediate production-grade substitutes. This also discounts the response of software businesses as they adapt strategy to become more relevant to their customers.

These misunderstandings can affect investment outcomes. For example, the sell-off exposed portfolios with a higher concentration in specific sectors or business types that are negatively impacted by the perception of AI disruption.

Portfolio implications

As always, diversifying alpha sources and their time horizon delivery will help build portfolio resilience. For multi-manager portfolios, that means diversifying capital thoughtfully across different investment styles to reflect portfolio objectives and active risk tolerance. Additionally, it is sensible to increase the role of core strategies to help navigate both sector-specific shocks and regime changes.

A sharp focus on active management and research-driven stock selection is also necessary to help navigate volatility and return dispersion.

Beyond these immediate implications for equity portfolios, there are second-order effects that arise from AI adoption to consider. These include the impact on the labour market – across both the technology sector and the broad white-collar workforce – as it could drive a potential decline in consumption growth.

The rout in the software sector has not been contained only to listed equities: credit and fixed interest markets are also affected. Investment-grade bonds have limited risk, while leveraged loans and private credit are more vulnerable due to the large number of software firms backed by private equity.

The impact on private equity sub-strategies also varies, with some areas of venture capital focused on software businesses likely to face valuation pressures.

These distinctions are critical for portfolio risk assessment and manager selection in an AI-disrupted environment.

The traditional valuation frameworks based on using prior selloffs as an anchor do not apply cleanly in the current environment, as AI may structurally reduce the market size of some segments of the software industry.

Investors in both listed equities and private markets must be vigilant against complacency and should understand how investment managers are evolving in a period of heightened disruption. Incorporating current market realities into investment processes is likely to help avoid costly valuation traps and instead allow investors to capitalise on the beneficiaries of AI that are beginning to emerge.

Matthew Gadsden is head of research execution at JANA Investment Advisers.